C AHIERS DE
TOPOLOGIE ET GÉOMÉTRIE DIFFÉRENTIELLE CATÉGORIQUES
T IMOTHY P ORTER
Abstract homotopy theory in procategories
Cahiers de topologie et géométrie différentielle catégoriques, tome 17, n
o2 (1976), p. 113-124
<
http://www.numdam.org/item?id=CTGDC_1976__17_2_113_0>
© Andrée C. Ehresmann et les auteurs, 1976, tous droits réservés.
L’accès aux archives de la revue « Cahiers de topologie et géométrie différentielle catégoriques » implique l’accord avec les conditions générales d’utilisation (
http://www.numdam.org/conditions). Toute utilisation commerciale ou impression systématique est constitutive d’une infraction pénale. Toute copie ou impression de ce fichier doit contenir la présente mention de copyright.
Article numérisé dans le cadre du programme
ABSTRACT HOMOTOPY THEORY IN PROCATEGORIES
by Timothy
PORTERCAHIERS DE TOPOLOGIE
ET GEOMETRIE DIFFERENTIELLE
Vol. XVII - 2 ( 1976 )
Artin and Mazur
in [1]
initiated the use of the category ofprosim- plicial
sets as a tool inhomotopy theory.
Their method involvedfirstly
theformation of the
homotopy
category ofsimplicial
sets and thenpassing
tothe
corresponding
procategory. This method proves to be unsuitable for cer-tain
applications
andby analogy
with a similar situation withcategories
ofsimplicial
spectra, it seems to be advisable to reverse the order of construc-tion. This was done in
[9]
and it allows one to use thehomotopy
limit con-struction of Bousfield and
Kan [3]y
or Boardman andVogt ([2] and [11]).
In [9]
no attempt was made toinvestigate
otherhomotopy
structures within the procategory.In this paper an attempt is made to extend a sizeable amount of « ab-
stract
homotopy theory»
from acategory
to thecorresponding
category ofproobjects.
Themeaning
we attach to the term «abstracthomotopy theory»
is that
of Quillen [10J
or Brown[4] ;
weonly
claim« attempt » because
theactual result obtained shows
that,
onextending
an abstracthomotopy
theo-ry in the manner
shown,
some of the structure is weakened.However,
it willbe shown that sufficient structure remains to
give
a sizeable amount of elem-entary
homotopy theory.
In future papersparticular
cases will be examinedin more detail and it will be shown
just
how much structure is retained.Other
extensions arepossible;
for instance seeHastings [8] ,
Ed-wards and
Hastings [6]
and Grossman[71 .
These structures retain moreof the
original homotopy theory on C
but do not allow the same sort of ap-plications.
Finally
I would like to thank Professor Wall ofLiverpool
and GavinWraith of Sussex for
indirectly suggesting
thisapproach.
1. Categories
offibrant objects.
In
[4],
Brown showed that a weaker form ofhomotopy theory
thanQuillen’s
«model category »structure [10] gives
a sufficient amount of stru- cture to allow a considerable part ofhomotopy theory
to be carried out. Ex-plicitly
he made thefollowing
definitions.Let
be a category with finiteproducts
and a finalobject e .
As-sume
that
has twodistinguished
classes of maps called weakequival-
ences and
librations.
A map is called anaspherical
or trivialfibration
ifit is both a fibration and a weak
equivalence.
A
path space object
for anobject
B is anobject BI together
with mapswhere s is a weak
equivalence, ( do, d1)
is a fibration andC
will be called a categoryojjibrant objects (
for ahomotopy theory)
if this structure satisfies the axioms :
( F1)
Iff,
g are maps so thatg f
is defined and two off, g
andjg are
weak
equivalences,
so is the third.( F2 )
Thecomposite
of two fibrations is a fibration.Any isomorphism
is a fibration.
( F3)
Given adiagram
with v a
fibration,
the fibredproduct
A x B exists and theprojection
I)r.*C
A x B -> A is a fibration. If v is
aspherical,
so is pr .C
(F4)
For anyobject B ,
there exists at least onepath space /3
I (notnecessarily
functorial in B).
(F5)
For anyobject B ,
the map B -> e is a fibration.We refer the interested reader to Brown’s paper
[4 ]
for the detaileddevelopment
of this set of axioms. For the situation in which these axiomswill be used
here,
it is necessary to alter themslightly.
Wereplace (F4) by
an axiom which
requires
that at least one functorialpath
spaceobject
ex-ists ;
we will however continue to call this one( F4).
It is a well known result
(
see for instance Artin and Mazur[1]
orDuskin [5])
that any mapf : A - B
inpro (C)
can bereplaced
up to iso-morphism by
a «level map)), i. e. aproobject
in the category of mapsof e .
Because of this it
helps
to look at functorcategories Hom (g , e)
before con-sidering
the more difficult case ofpJIQ( C).
We shall of course assume thatI
issmall,
but noassumption
rieeds be made as to other structureof g.
TH EOREM 1. 1. Let
Hom (g,C)
be thecategory of functors from 9
to a cat-egory
o f fibrant objects
withfunctorial (F4).
ThenHom. (g, e)
is a cat-egory
o f fibrant objects.
P RO O F . We will
specify
therequired
structure but will leave it to the readerto
verify
that this structure works.- Weak
equivalences:
Letand
be two functors and
f: X -Y
a naturaltransformation; f
is a weakequiva-
lence if
each f ( i ): X(i) ->Y(i)
is a weakequivalence in(?.
- Fibrations : With
f : X - Y
asabove, f
is a fibration if eachf( i ) :
X( i) ->
Y( i )
is a fibrationin e.
- Path space
object : Denoting
the functorialpath
spaceby ( )I ,
thepath
space ofX : I -+ e
isCOROLLARY 1. 2. The class
o f weak equivalences
inRam (g, e)
admits acalculus
of right fractions.
P RO O F . This is a direct consequence of the Theorem
together
withPropos-
ition
2,
page 424 of[4 ] .
The
corollary implies
that the category of fractions formedby
form-ally inverting
the weakequivalences
is well behaved. Acomparison
withthe work of
Vogt [11]
suggests that this«homotopy
category » isequivalent
to some category of coherent
9-indexed diagrams
in some sense ; we will notpursue this idea here.
2. Categories of cofibrant objects.
In
[4] ,
Brownonly gives explicitly
the axioms listed in n°1 ;
how-ever on page 442 he mentions the duals of these axioms in
Proposition
8 .Although
it is notabsolutely
necessary, we list below the dual axioms andstate the duals of Theorem 1.1 and
Corollary
1.2. The reason forstating
these axioms for a « category of ’cofibrant
objects »
is that anyQuillen
modelcategory [10] yields
both a category of fibrantobjects
and a category of cofibrantobjects.
It is this situation which is the mostimportant
one fromthe
point
of view of theapplications.
Let
be a category with finitecoproducts
and an initialobject e .
We assume
that
has twodistinguished
classes of maps called weak;equi-
valences andco fibrations.
A map in both classes is called a trivialco17*-
bration.
A
cylinder object
for anobject
B is anobject
B X Itogether
withmaps
with
o-(60+61)= v B,
thecodiagonal
map,60+61
acofibration,
and cra weak
equivalence.
C
will be called acategory of cofibrant objects ( for
ahomotopy theory)
if this structure satisfies the axioms :(C1)
as(F1) ,
(C2)
as( F2 )
with « f ibration »replaced by « cofibration » ( C3)
Given adiagram
with v a
cofibration,
thepushout A V B
exists andv’: A - A V B
is a co-C C
fibration which is trivial if v is.
(C4)
For eachB ,
there isat least
acylinder object
BxIin C
( whichwe will need to be functorial in B
).
( C5 )
For anyobject B ,
themap e - B
is a cofibration.EX A MP L E S.
( a) If C
is a model category in the sense ofQuillen [10]
andC
consists of the cofibrantobjects,
thenC c
is a category of cofibrant ob-jects
in the above sense.(b) If C
is a category of fibrantobjects,
theneop
is a category of co- fibrantobjects
in a.natural way.REMARK.
Although
we will not state or prove any of the dualresults,
thetheorems and
proofs
of Brown’s paper[4] easily
dualize and so we shallfeel free to use any of these dual results without
explicit proof
in the rem-ainder of this paper or in
sequels
to this paper.THEOREM 2. 1. Let
Hom. (g,C) be tbe category o f functors from I
to a cat-egory
of cofibrant objects e,
withfunctorial (C4).
ThenHom(g , C)
is acategory of co f i brant objects.
PROOF. As before the bulk of the work is left to the reader. If we define weak
equivalences
as in 1.1 and cofibrations in like manner, the result fol- lowseasily.
COROLLARY 2.2. The class
of
weakequivalences
inHom.( g, C)
admits acalculus of left fractions.
COROLL ARY 2.3.
If Hom (g, C) is
thecategory of functors from I
to a clo-sed model
category C,
thenJGm( g, ee) is a category o f fibrant objects
and
Hom (g, Cc.)
is acategory o f cofibrant objects. If C= Cc =Cf’ then
Hom. (g, C)
has both structures.RE M A R K S.
( a )
In2.3, Cf is, following Quillen’s
notation[10],
the full ’subcategory of e
determinedby
the fibrantobjects of e .
(b) Although Hom (g,C)
is both a category of fibrantand
of cofibrantobjects,
this does not mean it is aQuillen
model category. InQuillen’s
ax-ioms, (M1) requires
a certain interaction between fibrations andcofibrations;
there seems no reason to suppose that fibrations and cofibrations interact in this way in
Hom. (g,C).
It is for this reason thatcategories
of fibrantand cofibrant
objects
suggest themselves as suitable for functorcategories.
( c )
Given more structurein C, Hastings
hasequipped Hom (g, C)
witha full
Quillen
model category structure( see [8] ).
His definitions howeverare not suitable for the
applications envisaged
for thetheory being
develo-ped
here.3. Procotegori es.
As was mentioned
before,
any map in a procategorypta( ej
can bereindexed to
give
a «level map », and so ife
has ahomotopy
structure in-volving
weakequivalences,
etc..., it is natural to definef : X - Y
to be aweak
equivalence
inpro( C) if, by reindexing,
one gets a level mapfg:
Xi - Yi
indexedby
somecofiltering category g ,
which is a weakequiva-
lence in the
homotopy
structure ofHom (g, C),
andsimilarly
for fibrationsand cofibrations if any. From another
point
ofview,
thisapproach
to defin-ing
ahomotopy theory on pro (C) might
be considered as an attempt to«glue»
the various
homotopy
structures in the functorcategories Hom (g, C),
forcofiltering g, together
via final( or coinitial)
functors.Unfortunately
thisnaive
approach fails ;
for instance the class of weakequivalences
definedin this way is not closed under
composition.
The obvious way is to « generates an abstract
homotopy
from the « ba-sic» material
given by
theHom (g, C) .
We will thus make thefollowing
de-finitions :
- Basic weak
equivalence: f: X -Y
is a basic weakequivalence
ifby reindexing
one obtains a level weakequivalence,
i. e. a weakequival-
ence in the
image
of someHom(g, C) with g cofiltering. f
will also be called a basic weakequivalence
if it is anisomorphism.
- Basic
fibration ( or co fibration ) : f: X -
Y is a basic fibration if itcan be
replaced (by reindexing) by
a level fibration(similarly
forcofibration).
- Basic trivial
fibration :
as above but with a level trivial fibration.- Path
space :
IfX: I
->e
then as beforeX,
is thecomposite
-
Cylinder object :
IfX : I - e,
then :- Weak
equivalence : f : X
-Y is a weakequivalence
if it is a com-posite
of basic weakequivalences.
- Fibration : f : X -> Y is a fibration if it is a
composite
of basic fibra- tions andisomorphisms.
Even with these
definitions, pro (C)
is not a category of fibrant(or cof ibrant ) objects,
butby
careful use of thereindexing
results( [1] Appen- dix)
available inpro (C) ,
it ispossible
to recover the mostimportant
re-sults of Part 1 of Brown’s paper
[4] .
We will not prove all the intermediate steps in theproof
of these moregeneral results,
butmerely
indicate how toadapt
theproofs given in [4] .
First note that there is a weak form of axiom system satisfied
by
the classes described above. The
important
differences are that the map(do d 1)
is a basic fibration and in axiom(F3)
we canonly
claimthat,
ifv is a basic trivial
fibration,
then so is pr( similarly
for( C3 ) ) .
LEMMA
3.1. (Factorization
lemma).1 f
u is anymap
inpro (C),
then u canbe
factored
u =p i ,
where p is a basicfibration
and i isright
inverse to abasic trivial
fibration.
PROOF.
Represent
uby
a level map andapply
the factorization lemma in the relevantHom (g, C) .
Th e definition of
ohomotopic*
isinteresting
in its ownright
and in-dicates the sort of structure that is involved here.
If
f, g : A -> B
are two maps inpro (C)
thenthey
arebomotopic
ifthere is a
« homotopy s
such that and
We will write
There is
immediately
theproblem
ofproving = >>
is transitive.Wecan
reinterprete =>> by noting
that f=
g if andonly
ifby reindexing
onegets a
diagram
so that in
LEMMA 3.2 and th en
P ROO F.
Suppose fg = gg
andgi, = hi, ;
then there is a smallcofiltering
cat- -egory with
objects pairs (i, j)
with iin g, j E g’
such that there is a dia- gramrepresenting
thediagram
(with
the vertical maps therespective identities)
inpro (C).
A map from( i , j )
to(i’, j’)
is a map of thecorresponding diagrams.
If we denote thissmall
cofiltering
categoryby 11
we get aprodiagram by assigning
to each( i , j )
thecorresponding diagram *( i , j ) .
Each
diagram
thus formedsimplifies
togive
and on
noting that
we canjoin
the twohomotopies
t ogether
togive
ahomotopy
and this is the
required homotopy,
f= h .
LEMMA
3.3 (cf.
Brown[4] Lemma 1). Any diagram
with i a basic weak
equivalence
and p afibration
can be imbedded( up
toisomorphism)
in adiagram
with t a
composite of
basic trivialfibrations.
P RO O F . First
apply
in case p is a basic fibration and thenusing
induc-tion on the
length
of thecomposite. Again reindexing
is used to enable eachstage to be done in a functor category.
It is
only
this weaker form of Lemma 3.3 which isneeded,
one neverneeds the stronger form where i is a weak
equivalence.
Brown’s Lemma 2 and
Propositions
1 and 2 now followeasily using
induction on the
lengths
ofcomposite required
in the weakequivalences.
As a result of
this,
Theorem 1 page425
holds and we can form the homo- topy categoryhapna C) using
(where
the limit is over the category of weakequivalences
overA )
as thehom-set.
Rather than dwell too
long
on the minorchanges
needed in Brown’ssubsequent proofs
to effect the extension topro (C) -basically
thechanges
are
managed by
carefulreindexing along
the lines of Lemma3.2
above andinduction on the
lengths
of fibrations and weakequivalences -
we will leavethe statement and
proofs
of intermediate results to the reader and will mere-ly
state the theorem which is most useful in theapplications.
We first need a definition from
[4]
page 432for C pointed.
Afibra-
tion sequence is a
diagram F - E -
B inC) together
with an actionin
Hopro (C), F X D B - F,
which isisomorphic
to adiagram
and action ob-tained from a fibration in pro
(C) .
TH E O R E M 3.4.
If C
is apointed category o f fibrant objects and .0
denotesthe
loop space functor,
thenfor
anyfibration
sequencein
p.tu1( C) ,
the sequenceis
exact (
in the senseof Quillen [10] for
q =1 ) for
any A in pro( e) . Dualising
the whole of this section we get the dual result.THEOREM 3.5.
If e
is apointed category o f cofibrant objects and E
den-otes the
suspension functor,
thenfor
anycofibration
sequencein
pm( e)
the sequenceis exact
(
asbefore) for
any B i npro ( C) -
The
interesting
morespecial
case iswhere
is apointed Quillen
model category and as before
Cf
is the category of fibrantobjects
andC
the category of cofibrant
objects
ine.
Theorems3.4
and3.5 apply
inpro (Cf)
and
pro ( ee) respectively
and theA
of 3.4( and
B of3.5)
can be chosenfrom amongst those
objects weakly equivalent
to someobject
inpro (Cf)
(respectively
inpro (Cc )).
As mentioned in Section
2,
no such axiom asQuillen’s (M1)
can be satisfied inpro (C)
evenif ee = Cf, but by using
a stronger form of
fibration or cofibration such a
development
ispossible ( see [6] or [8] ).
If,
inaddition, () I
and() X I
areadjoint. on (C
there is somelinking
bet-ween the two forms of structure
( again assuming
forsimplicity C = Cc = (Cf).
In the cases
(i) C = Kano,
the category ofpointed
Kancomplexes,
and
(ii) C = C+(Mod-A),
the category of chaincomplexes
of A-moduleswhich are bounded
below,
such a
linking
occurs. We willinvestigate
those two cases in future papers.Department of Mathematics University College CORK
Republic of IRELAND
REFERENCES
1. ARTIN, M., MAZUR, B., Etale homotopy, Lecture
Notes in
Math. 100, Springer (1969).2. BOARDMAN, J.M., VOGT, R.M., Homotopy invariant algebraic structures on top-
ological spaces, Lecture Notes in Math. 347, Springer ( 1973).
3. BOUSFIELD, A.K., KAN, D.M., Homotopy limits, completions and localizations, Lecture Notes in Math. 304, Springer ( 1972 ).
4. BROWN, K.S., Abstract homotopy theory and generalized sheaf cohomology, Trans.
A. M. S. 186 (1973), 419-458.
5. DUSKIN, J., Pro-objects, Séminaire
Heidelberg-Strasbourg
1966-67, Exposé 6, I. R. M. A. Strasbourg.6. EDWARDS, D.A., HASTINGS, H.M., On
topological
methods inhomological algebra,
preprint, S. U. N. Y., Binghamton.
7. GROSSMAN, J.W., A homotopy theory for pro-spaces, Trans. A. M. S. 201 (1975), 161-176.
8. HASTINGS, H.M.,
Homotopy theory of
prospaces I-III, preprints, S. U. N. Y. Bin-gh am ton.
9. PORTER, T., Stability results for topological spaces, Math. Z. 140 ( 1974), 1-21.
10. QUILLEN, D., Homotopical algebra, Lecture Notes in Math. 43, Springer (1967).
11. VOGT, R.M., Homotopy limits and colimits, Math. Z. 134 ( 1973), 11-52.